Coinbase announced that its artificial intelligence-driven security systems intercepted $4.7 million in attempted transfers to attackers who targeted viewers with a deepfake of Elon Musk during a hijacked YouTube livestream of Tesla’s “flying car” demonstration. The fraudulent event drew more than 500,000 viewers, many of whom were enticed to send cryptocurrency by promises that their funds would be doubled if they used a QR code flashed on screen.
AI defenses halt crypto scam attempts
Within approximately two hours, hundreds of Coinbase users attempted to send millions of dollars to the scammers’ wallets. According to the exchange, its automated systems blocked around 91% of these transactions before any funds could be delivered, without requiring intervention from employees.
Scammers managed to lure a livestream audience of more than 500,000 people with a fake Tesla crypto giveaway promoted by an AI-generated image of Elon Musk. In about two hours, hundreds of Coinbase customers tried sending millions of dollars, but Coinbase’s AI-powered scam defenses were able to stop $4.7 million from reaching the attackers.
The exchange reported that the attack relied on deepfake technology, a method that uses artificial intelligence to convincingly create images, audio, or video of public figures. In this case, the attackers mimicked Elon Musk, CEO of Tesla, to lend credibility to the fake offer and drive engagement among cryptocurrency users.
Mini dictionary: Deepfake, a term for synthetic media in which artificial intelligence and machine learning are used to generate realistic audio, video, or images that can impersonate people, often with deceptive intent.
Record levels of crypto theft in 2026
The use of AI technology in scams comes during a record-breaking year for crypto-related theft. According to DeFiLlama, there have been 343 major attacks since the start of the year, resulting in $8.78 billion in stolen assets. One of the most significant incidents includes October’s Binance breach, which has been attributed to TraderTraitor, an alleged North Korean hacking group linked to the largest theft in cryptocurrency history.
| Metric | 2026 (Year-to-date) | Main Contributor |
|---|---|---|
| Total crypto thefts | 343 incidents | Binance hack |
| Total value stolen | $8.78 billion | TraderTraitor group |
While high-profile hacking incidents dominate the headlines, scams are growing at an even faster pace. TRM Labs reported in August that losses resulting from deepfake scams in 2026 surged by 263% compared to the total reported in 2025. Security specialists note that these scams require less technical skill, as they trick users into willingly transferring funds rather than infiltrating technical systems.
Mini dictionary: TRM Labs, a blockchain intelligence company that provides compliance and risk management solutions to detect fraud and financial crime in digital assets.
Evolving threat and improved defenses
The most recent scam echoed a similar incident that took place in July 2020. At that time, several prominent Twitter accounts, including Elon Musk’s, were compromised, and a comparable offer claimed users’ funds would be doubled if sent to a listed address. Coinbase responded by manually blacklisting the scammers’ Bitcoin address, though this only happened once payments were already underway.
Six years later, while the core scam has changed little, the technology has advanced significantly. The transition from a compromised tweet to a highly convincing AI-generated livestream has exponentially increased the potential reach, bringing hundreds of thousands directly into contact with the fraud.
Coinbase has adapted its approach in response. The platform’s AI-driven system now steps in to prevent customers from making costly errors, acting automatically before funds leave user accounts. With deepfakes becoming more sophisticated and widespread, Coinbase has shifted its security focus from solely technical exploits to addressing social engineering schemes as well.
The exchange’s evolving protective measures recognize that user mistakes, especially when exploited through realistic AI-generated content, are no longer seen only as individual responsibility but as a risk to be actively mitigated by the platform’s security infrastructure.






